Prepare a client meeting with Sellsy, from Claude or ChatGPT
One question to your AI, and a client’s Sellsy record, conversation history and opportunities become a meeting brief. With your Sellsy permissions, no export.
Published on October 28, 2026
In short
- Connect Sellsy to your assistant through Persyst: it reads a client’s company record, timeline and opportunities, then writes the meeting brief.
- The assistant reads Sellsy live, with the Sellsy permissions of the connected person. Nothing is copied or written to the CRM.
- One question is enough: “Prepare my meeting with Dupont SA”. The assistant chains the search, the reads and the summary.
Prepare my meeting with Dupont SA tomorrow: where do we stand, what is blocking, and what should I offer?
Dupont SA has been a client since 2024. Open opportunity “2027 renewal”, €24,000, Negotiation stage, owned by Camille. Last exchange on October 2: Marc Dupont asks for a volume discount before signing. Watch out: a quote from June never got an answer. To offer: the volume discount in exchange for a two-year commitment, and an answer on the June quote.
Sources consulted
- Sellsy · company record
- Sellsy · timeline
- Sellsy · opportunities and quotes
Why meeting prep takes time
Before a meeting, a sales rep opens the company record, scrolls the timeline, finds the latest opportunity, checks pending quotes and notes who will attend. The information is in Sellsy, but spread across several screens. The summary still has to be written.
An AI assistant can write that summary. What it lacks is access to the CRM, with the right permissions.
What the assistant does, step by step
With Persyst’s Sellsy connector, the question “Prepare my meeting with Dupont SA” triggers four reads:
- Search for the company in the CRM (
search_sellsy_crm). When several match, the assistant shows the candidates before going further. - Read the company record and its contacts (
get_sellsy_record). - History of exchanges, notes and events (
search_sellsy_timeline). - Linked opportunities and quotes (
get_sellsy_object), with the pipeline stages to interpret them (get_sellsy_reference).
The assistant then cross-checks these elements and writes: where the relationship stands, what is blocking, what to offer. The sources it consulted are listed under the answer.
Three questions to try
- “Summarize the last three exchanges with Dupont SA and list the commitments made on each side.”
- “Which quotes sent to Dupont SA are still unanswered?”
- “Who are the active contacts at Dupont SA, and who was contacted last?”
Setup
- An administrator creates the company’s Sellsy API access and enters it in Persyst. The Sellsy guide walks through each screen.
- Each sales rep connects their Sellsy account in My connectors: Sellsy shows its consent screen, no password is entered in Persyst.
- They add Persyst as an MCP server in their assistant (getting started), then ask their question.
An administrator chooses which Sellsy tools each team can use, and Persyst’s history keeps who read what.
Limits to know
- The assistant only reads what Sellsy allows the connected sales rep to see.
- Invoices, the product catalog and the Files module are not covered by the current tools.
- Nothing is written to Sellsy: the meeting notes still go into the CRM.
Questions
Frequently asked questions
Can the assistant update the Sellsy record after the meeting?
No. Persyst’s Sellsy connector is read-only: no record, note or opportunity is created or changed. The meeting notes are entered in Sellsy, as today.
Does a sales rep see their colleagues’ clients?
Only if Sellsy already allows it. Each member connects their own Sellsy account; Sellsy’s privilege profiles and sharing apply to every read by the assistant.
Do I need to export the CRM or hand the AI a file?
No. The assistant queries Sellsy at the time of the question, through Persyst’s tools. There is no export, no copy and no intermediate database to keep up to date.
Which assistants can prepare this meeting?
Claude, ChatGPT, Copilot, Le Chat or any MCP-compatible assistant in which the member has added Persyst. The result depends on the chosen model; the data read is the same.